Industrial Alarm Management using Reinforcement Learning

使用强化学习的工业警报管理

• 批准号: 519806-2017
• 负责人: Liu, Jinfeng
• 金额: $ 1.82万
• 依托单位: University of Alberta
• 依托单位国家: 加拿大
• 项目类别: Engage Grants Program
• 财政年份: 2017
• 资助国家: 加拿大
• 起止时间: 2017-01-01 至 2018-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=640592
• 关键词: Industrial; Alarm; Management; using; Reinforcement

Abnormal situations cost industry millions of dollars every year. Alarm systems play important roles inensuring safe and efficient operation of modern industrial plants. However, most existing industrial alarmsystems suffer from alarm overloading. This implies that operators have to deal with too many alarms thatexceed the operators' capacity. Typically, plant operators receive much more alarms than they can handle.There has been an increasing interest in industry to address this issue and seek remedies to reduce the numberof false and nuisance alarms. Moreover, the current alarm messages do not indicate the root causes of thealarms or how to solve the alarms. An effective root cause analysis is highly desired.NTWIST Inc. provides advanced real-time alarm management solutions to process industries. Their goal is toeliminate preventable valve closures, unplanned equipment shutdowns and unplanned facilityoutages/shutdowns caused by alarm flooding and alarm suppression. Currently, NTWIST is developing areal-time alarm management system based on artificial intelligence. In this project, NTWIST works togetherwith Dr. Liu's group in the University of Alberta to learn from historical alarm data using reinforcementlearning and deep learning to address the issues with alarm flooding based on each alarm's impact on theprocess and safety conditions.

异常情况每年使工业损失数百万美元。报警系统在保证现代工业设备安全、高效运行中起着重要的作用。然而，大多数现有的工业报警系统遭受报警过载。这意味着操作员必须处理超过操作员能力的太多警报。通常情况下，工厂操作员收到的警报比他们能够处理的要多得多。工业界对解决这个问题越来越感兴趣，并寻求补救措施以减少误报和骚扰警报的数量。此外，目前的报警信息没有指出报警的根本原因或如何解决报警。一个有效的根本原因分析是非常需要的。为流程工业提供先进的实时报警管理解决方案。他们的目标是消除可预防的阀门关闭、非计划的设备关闭和非计划的设施中断/关闭，这些都是由报警溢流和报警抑制引起的。目前，公司正在开发基于人工智能的实时报警管理系统。在这个项目中，NTWIST与阿尔伯塔大学的刘博士团队合作，使用协同学习和深度学习从历史警报数据中学习，根据每个警报对过程和安全条件的影响来解决警报泛滥问题。